# Building a Strong Instruction Language Model for a Less-Resourced Language

arXiv: https://arxiv.org/abs/2603.01691 · HTML: https://mlsanity.com/paper/2603.01691/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 44 references that are in the map.
- First posted on arXiv 2 Mar 2026.

## Builds on (most cited first, 22 of 22)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline](https://mlsanity.com/paper/2406.11939/) (2024)
- [LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset](https://mlsanity.com/paper/2309.11998/) (2023)
- [Efficient and Effective Text Encoding for Chinese LLaMA and Alpaca](https://mlsanity.com/paper/2304.08177/) (2023)
- [Learning Word Vectors for 157 Languages](https://mlsanity.com/paper/1802.06893/) (2018)
- [CometKiwi: IST-Unbabel 2022 Submission for the Quality Estimation Shared Task](https://mlsanity.com/paper/2209.06243/) (2022)
- [FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language](https://mlsanity.com/paper/2506.20920/) (2025)
- [NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model](https://mlsanity.com/paper/2508.14444/) (2025)
- [EuroLLM: Multilingual Language Models for Europe](https://mlsanity.com/paper/2409.16235/) (2024)
- [Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance](https://mlsanity.com/paper/2507.22448/) (2025)
- [Improving In-context Learning of Multilingual Generative Language Models with Cross-lingual Alignment](https://mlsanity.com/paper/2311.08089/) (2023)
- [Recycling the Web: A Method to Enhance Pre-training Data Quality and Quantity for Language Models](https://mlsanity.com/paper/2506.04689/) (2025)
- [FinEst BERT and CroSloEngual BERT: less is more in multilingual models](https://mlsanity.com/paper/2006.07890/) (2020)
- [Advancing Generative AI for Portuguese with Open Decoder Gerv'asio PT*](https://mlsanity.com/paper/2402.18766/) (2024)
- [Sequence to sequence pretraining for a less-resourced Slovenian language](https://mlsanity.com/paper/2207.13988/) (2022)
- [Generative Model for Less-Resourced Language with 1 billion parameters](https://mlsanity.com/paper/2410.06898/) (2024)
- [CLASSLA-web: Comparable Web Corpora of South Slavic Languages Enriched with Linguistic and Genre Annotation](https://mlsanity.com/paper/2403.12721/) (2024)
- [Improving LLMs for Machine Translation Using Synthetic Preference Data](https://mlsanity.com/paper/2508.14951/) (2025)

## Built on by (most cited first, 1 of 1)

- [Maistros: A Greek Large Language Model Adapted Through Knowledge Distillation From Large Reasoning Models](https://mlsanity.com/paper/2605.01870/) (2026)

## Abstract (arXiv)

Large language models (LLMs) have become an essential tool for natural language processing and artificial intelligence in general. Current open-source models are primarily trained on English texts, resulting in poorer performance on less-resourced languages and cultures. We present a set of methodological approaches necessary for the successful adaptation of an LLM to a less-resourced language, and demonstrate them using the Slovene language. We present GaMS3-12B, a generative model for Slovene with 12 billion parameters, and demonstrate that it is the best-performing open-source model for Slovene within its parameter range. We adapted the model to the Slovene language using three-stage continual pre-training of the Gemma 3 model, followed by two-stage supervised fine-tuning (SFT). We trained the model on a combination of 140B Slovene, English, Bosnian, Serbian, and Croatian pretraining tokens, and over 200 thousand English and Slovene SFT examples. We evaluate GaMS3-12B on the Slovenian-LLM-Eval datasets, English-to-Slovene translation, and the Slovene LLM arena. We show that the described model outperforms 12B Gemma 3 across all three scenarios and performs comparably to much larger commercial GPT-4o in the Slovene LLM arena, achieving a win rate of over 60 %.